Digital video consumes large amounts of storage and transmission capacity. A typical raw digital video sequence includes 15 or 30 frames per second. Each frame can include tens or hundreds of thousands of pixels (also called pets). Each pixel represents a tiny element of the picture. In raw form, a computer commonly represents a pixel with 24 bits. Thus, the number of bits per second, or bitrate, of a typical raw digital video sequence can be 5 million bits/second or more.
Most computers and computer networks lack the resources to process raw digital video. For this reason, engineers use compression (also called coding or encoding) to reduce the bitrate of digital video. Compression can be lossless, in which quality of the video does not suffer but decreases in bitrate are limited by the complexity of the video. Or, compression can be lossy, in which quality of the video suffers but decreases in bitrate are more dramatic. Decompression reverses compression.
In general, video compression techniques include intraframe compression and interframe compression. Intraframe compression techniques compress individual frames, typically called I-frames, or key frames. Interframe compression techniques compress frames with reference to preceding and/or following frames, and are called typically called predicted frames, P-frames, or B-frames.
Microsoft Corporation's Windows Media Video, Version 7 [“WMV7”] includes a video encoder and a video decoder. The WMV7 encoder uses intraframe and interframe compression, and the WMV7 decoder uses intraframe and interframe decompression.
A. Intraframe Compression in WMV7
FIG. 1 illustrates block-based intraframe compression (100) of a block (105) of pixels in a key frame in the WMV7 encoder. A block is a set of pixels, for example, an 8×8 arrangement of pixels. The WMV7 encoder splits a key video frame into 8×8 blocks of pixels and applies an 8×8 Discrete Cosine Transform [“DCT”] (110) to individual blocks such as the block (105). A DCT is a type of frequency transform that converts the 8×8 block of pixels (spatial information) into an 8×8 block of DCT coefficients (115), which are frequency information. The DCT operation itself is lossless or nearly lossless. Compared to the original pixel values, however, the DCT coefficients are more efficient for the encoder to compress since most of the significant information is concentrated in low frequency coefficients (conventionally, the upper left of the block (115)) and many of the high frequency coefficients (conventionally, the lower right of the block (115)) have values of zero or close to zero.
The encoder then quantizes (120) the DCT coefficients, resulting in an 8×8 block of quantized DCT coefficients (125). For example, the encoder applies a uniform, scalar quantization step size to each coefficient, which is analogous to dividing each coefficient by the same value and rounding. For example, if a DCT coefficient value is 163 and the step size is 10, the quantized DCT coefficient value is 16. Quantization is lossy. The reconstructed DCT coefficient value will be 160, not 163. Since low frequency DCT coefficients tend to have higher values, quantization results in loss of precision but not complete loss of the information for the coefficients. On the other hand, since high frequency DCT coefficients tend to have values of zero or close to zero, quantization of the high frequency coefficients typically results in contiguous regions of zero values. In addition, in some cases high frequency DCT coefficients are quantized more coarsely than low frequency DCT coefficients, resulting in greater loss of precision/information for the high frequency DCT coefficients.
The encoder then prepares the 8×8 block of quantized DCT coefficients (125) for entropy encoding, which is a form of lossless compression. The exact type of entropy encoding can vary depending on whether a coefficient is a DC coefficient (lowest frequency), an AC coefficient (other frequencies) in the top row or left column, or another AC coefficient.
The encoder encodes the DC coefficient (126) as a differential from the DC coefficient (136) of a neighboring 8×8 block, which is a previously encoded neighbor (e.g., top or left) of the block being encoded, (FIG. 1 shows a neighbor block (135) that is situated to the left of the block being encoded in the frame.) The encoder entropy encodes (140) the differential.
The entropy encoder can encode the left column or top row of AC coefficients as a differential from a corresponding column or row of the neighboring 8×8 block. FIG. 1 shows the left column (127) of AC coefficients encoded as a differential (147) from the left column (137) of the neighboring (to the left) block (135). The differential coding increases the chance that the differential coefficients have zero values. The remaining AC coefficients are from the block (125) of quantized DCT coefficients.
The encoder scans (150) the 8×8 block (145) of predicted, quantized AC DCT coefficients into a one-dimensional array (155) and then entropy encodes the scanned AC coefficients using a variation of run length coding (160). The encoder selects an entropy code from one or more run/level/last tables (165) and outputs the entropy code.
A key frame contributes much more to bitrate than a predicted frame. In low or mid-bitrate applications, key frames are often critical bottlenecks for performance, so efficient compression of key frames is critical.
FIG. 2 illustrates a disadvantage of intraframe compression such as shown in FIG. 1 In particular, exploitation of redundancy between blocks of the key frame is limited to prediction of a subset of frequency coefficients (e.g., the DC coefficient and the left column (or top row) of AC coefficients) from the left (220) or top (230) neighboring block of a block (210). The DC coefficient represents the average of the block, the left column of AC coefficients represents the averages of the rows of a block, and the top row represents the averages of the columns. In effect, prediction of DC and AC coefficients as in WMV7 limits extrapolation to the row-wise (or column-wise) average signals of the left (or top) neighboring block. For a particular row (221) in the left block (220), the AC coefficients in the left DCT coefficient column for the left block (220) are used to predict the entire corresponding row (211) of the block (210). The disadvantages of this prediction include:
1) Since the prediction is based on averages, the far edge of the neighboring block has the same influence on the predictor as the adjacent edge of the neighboring block, whereas intuitively the far edge should have a smaller influence.
2) Only the average pixel value across the row (or column) is extrapolated.
3) Diagonally oriented edges or lines that propagate from either predicting block (top or left) to the current block are not predicted adequately.
4) When the predicting block is to the left, there is no enforcement of continuity between the last row of the top block and the first row of the extrapolated block.
B. Interframe Compression in WMV7
Interframe compression in the WMV7 encoder uses block-based motion compensated prediction coding followed by transform coding of the residual error. FIGS. 3 and 4 illustrate the block-based interframe compression for a predicted frame in the WMV7 encoder. In particular, FIG. 3 illustrates motion estimation for a predicted frame (310) and FIG. 4 illustrates compression of a prediction residual for a motion-estimated block of a predicted frame.
The WMV7 encoder splits a predicted frame into 8×8 blocks of pixels. Groups of 4 8×8 blocks form macroblocks. For each macroblock, a motion estimation process is performed. The motion estimation approximates the motion of the macroblock of pixels relative to a reference frame, for example, a previously coded, preceding frame. In FIG. 3, the WMV7 encoder computes a motion vector for a macroblock (315) in the predicted frame (310). To compute the motion vector, the encoder searches in a search area (335) of a reference frame (330). Within the search area (335), the encoder compares the macroblock (315) from the predicted frame (310) to various candidate macroblocks in order to find a candidate macroblock that is a good match. The encoder can check candidate macroblocks every pixel or every ½ pixel in the search area (335), depending on the desired motion estimation resolution for the encoder. Other video encoders check at other increments, for example, every ¼ pixel. For a candidate macroblock, the encoder checks the difference between the macroblock (315) of the predicted frame (310) and the candidate macroblock and the cost of encoding the motion vector for that macroblock. After the encoder finds a good matching macroblock, the block matching process ends. The encoder outputs the motion vector (entropy coded) for the matching macroblock so the decoder can find the matching macroblock during decoding. When decoding the predicted frame (310), a decoder uses the motion vector to compute a prediction macroblock for the macroblock (315) using information from the reference frame (330). The prediction for the macroblock (315) is rarely perfect, so the encoder usually encodes 8×8 blocks of pixel differences (also called the error or residual blocks) between the prediction macroblock and the macroblock (315) itself.
Motion estimation and compensation are effective compression techniques, but various previous motion estimation/compensation techniques (as in WMV7 and elsewhere) have several disadvantages, including:
1) The resolution of the motion estimation (i.e., pixel, ½ pixel, ¼ pixel increments) does not adapt to the video source. For example, for different qualities of video source (clean vs. noisy), the video encoder uses the same resolution of motion estimation, which can hurt compression efficiency.2) For ¼ pixel motion estimation, the search strategy fails to adequately exploit previously completed computations to speed up searching.3) For ¼ pixel motion estimation, the search range is too large and inefficient. In particular, the horizontal resolution is the same as the vertical resolution in the search range, which does not match the motion characteristics of many video signals.4) For ¼ pixel motion estimation, the representation of motion vectors is inefficient to the extent bit allocation for horizontal movement is the same as bit allocation for vertical resolution.
FIG. 4 illustrates the computation and encoding of an error block (435) for a motion-estimated block in the WMV7 encoder. The error block (435) is the difference between the predicted block (415) and the original current block (425). The encoder applies a DCT (440) to error block (435), resulting in 8×8 block (445) of coefficients. Even more than was the case with DCT coefficients for pixel values, the significant information for the error block (435) is concentrated in low frequency coefficients (conventionally, the upper left of the block (445)) and many of the high frequency coefficients have values of zero or close to zero (conventionally, the lower right of the block (445)).
The encoder then quantizes (450) the DCT coefficients, resulting in an 8×8 block of quantized DCT coefficients (455). The quantization step size is adjustable. Again, since low frequency DCT coefficients tend to have higher values, quantization results in loss of precision, but not complete loss of the information for the coefficients. On the other hand, since high frequency DCT coefficients tend to have values of zero or close to zero, quantization of the high frequency coefficients results in contiguous regions of zero values. In addition, in some cases high frequency DCT coefficients are quantized more coarsely than low frequency DCT coefficients, resulting in greater loss of precision/information for the high frequency DCT coefficients.
The encoder then prepares the 8×8 block (455) of quantized DCT coefficients for entropy encoding. The encoder scans (460) the 8×8 block (455) into a one dimensional array (465) with 64 elements, such that coefficients are generally ordered from lowest frequency to highest frequency, which typical creates long runs of zero values.
The encoder entropy encodes the scanned coefficients using a variation of run length coding (470). The encoder selects an entropy code from one or more run/level/last tables (475) and outputs the entropy code.
FIG. 5 shows the decoding process (500) for an inter-coded block. Due to the quantization of the DCT coefficients, the reconstructed block (575) is not identical to the corresponding original block. The compression is lossy.
In summary of FIG. 5, a decoder decodes (510, 520) entropy-coded information representing a prediction residual using variable length decoding and one or more run/level/last tables (515). The decoder inverse scans (530) a one-dimensional array (525) storing the entropy-decoded information into a two-dimensional block (535). The decoder inverse quantizes and inverse discrete cosine transforms (together, 540) the data, resulting in a reconstructed error block (545). In a separate path, the decoder computes a predicted block (565) using motion vector information (555) for displacement from a reference frame. The decoder combines (570) the predicted block (555) with the reconstructed error block (545) to form the reconstructed block (575).
The amount of change between the original and reconstructed frame is termed the distortion and the number of bits required to code the frame is termed the rate. The amount of distortion is roughly inversely proportional to the rate. In other words, coding a frame with fewer bits (greater compression) will result in greater distortion and vice versa. One of the goals of a video compression scheme is to try to improve the rate-distortion—in other words to try to achieve the same distortion using fewer bits (or the same bits and lower distortion).
Compression of prediction residuals as in WMV7 can dramatically reduce bitrate while slightly or moderately affecting quality, but the compression technique is less than optimal in some circumstances. The size of the frequency transform is the size of the prediction residual block (e.g., an 8×8 DCT for an 8×8 prediction residual). In some circumstances, this fails to exploit localization of error within the prediction residual block.
C. Post-processing with a Deblocking Filter in WMV7
For block-based video compression and decompression, quantization and other lossy processing stages introduce distortion that commonly shows up as blocky artifacts—perceptible discontinuities between blocks.
To reduce the perceptibility of blocky artifacts, the WMV7 decoder can process reconstructed frames with a deblocking filter. The deblocking filter smoothes the boundaries between blocks.
While the deblocking filter in WMV7 improves perceived video quality, it has several disadvantages. For example, the smoothing occurs only on reconstructed output in the decoder. Therefore, prediction processes such as motion estimation cannot take advantage of the smoothing. Moreover, the smoothing by the post-processing filter can be too extreme.
D. Standards for Video Compression and Decompression
Aside from WMV7, several international standards relate to video compression and decompression. These standards include the Motion Picture Experts Group [“MPEG”] 1, 2, and 4 standards and the H.261, H.262, and H.263 standards from the International Telecommunication Union [“ITU”]. Like WMV7, these standards use a combination of intraframe and interframe compression, although the standards typically differ from WMV7 in the details of the compression techniques used. For additional detail about the standards, see the standards' specifications themselves.
Given the critical importance of video compression and decompression to digital video, it is not surprising that video compression and decompression are richly developed fields. Whatever the benefits of previous video compression and decompression techniques, however, they do not have the advantages of the following techniques and tools.